Method and device for generating site pollution treatment scheme

Abstract

The invention discloses a site pollution abatement scheme generation method and device, and the method comprises the steps: collecting the pollution data of a target region in multiple dimensions, and the pollution data at least comprise pollutant characteristic data, site characteristic data and environmental factor data; performing data processing on the pollution data of each dimension, and fusing the pollution data of each dimension after data processing to obtain fused multi-modal data; inputting the fused multi-modal data into a pre-constructed scheme generation model to obtain a pollution abatement scheme output by the scheme generation model; wherein the scheme generation model is constructed by utilizing the historical site pollution data subjected to data processing and a historical treatment scheme corresponding to the historical site pollution data. Through model construction and calling, the pollution control scheme for selection and reference is generated, and the technical problem that generation of the site pollution control scheme completely depends on artificial experience and site investigation is solved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method and device for generating a site pollution control plan. Background Art

[0002] Site pollution problems are usually related to industrial production, chemical waste leakage or historical pollution. Especially in the process of industrial sites, abandoned chemical plants and urban construction expansion, the demand for pollution control has increased significantly. Traditional site pollution control methods rely on expert experience and on-site investigations, which are not only time-consuming and labor-intensive, but also highly subjective. In addition, when dealing with complex pollutants, there are problems such as difficulty in accurate identification, insufficient monitoring data and inflexible control solutions.

[0003] In view of this, a method and device for generating a site pollution control plan are provided, in order to solve the technical problem that the generation of the site pollution control plan completely depends on manual experience and on-site investigation. Summary of the invention

[0004] To this end, the embodiments of the present invention provide a method and device for generating a site pollution control plan, which generates pollution control plans for selection and reference through model construction and retrieval, in order to solve the technical problem that the generation of site pollution control plans is completely dependent on manual experience and on-site investigation.

[0005] In order to achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0006] The present invention provides a method for generating a site pollution control plan, the method comprising:

[0007] Collecting pollution data of multiple dimensions in the target area, wherein the pollution data at least includes pollutant characteristic data, site characteristic data and environmental factor data;

[0008] The pollution data of each dimension are processed respectively, and the pollution data of each dimension after the data processing are fused to obtain fused multimodal data;

[0009] Inputting the fused multimodal data into a pre-built solution generation model to obtain a pollution control solution output by the solution generation model;

[0010] The solution generation model is constructed using various types of historical site pollution data after data processing, and historical governance solutions corresponding to the historical site pollution data.

[0011] In some embodiments, constructing a solution generation model specifically includes:

[0012] Obtain a large amount of historical site contamination data of various types;

[0013] Clean the historical site pollution data and mark the target information;

[0014] By using natural language processing algorithms, target features are extracted from the annotated historical site pollution data. Combined with semantic understanding and keyword matching, a large model of site pollution remediation technology based on retrieval enhanced generation is constructed.

[0015] In some embodiments, the pollutant characteristic data includes at least the type, concentration and spatial distribution of the pollutant; and / or,

[0016] The site characteristic data includes at least soil type, pH value, organic matter content and groundwater distribution; and / or,

[0017] The environmental factor data at least include temperature, humidity and rainfall data of the contaminated site.

[0018] In some embodiments, the pollutant characteristic data is processed, and then the process further includes:

[0019] Determine the concentration levels of target pollutants in soil and water samples based on extensive collection of soil and water samples in the target area;

[0020] The concentration distribution map of the target pollutant is drawn by using the sampling data and the spatial interpolation method, and the spatial distribution characteristics and migration trends of the target pollutant are determined based on the concentration distribution map.

[0021] In some embodiments, the data types of the pollution data include at least text data, image data, and sensor reading data.

[0022] In some embodiments, the pollution data of each dimension after data processing is fused, specifically including:

[0023] The text data is encoded using a word embedding algorithm to obtain an embedding vector corresponding to the historical governance record text data of the site to be governed;

[0024] extracting image features from the image data;

[0025] Extract pollutant and environmental numerical characteristics;

[0026] The embedding vector and the image feature are subjected to a data fusion algorithm to obtain a fusion feature output by the deep learning model.

[0027] In some embodiments, document feature extraction uses TF-IDF (Term Frequency-Inverse Document Frequency):

[0028]

[0029] Among them, TF(t,d) is the frequency of term t in document d, N is the total number of documents, and DF(t) is the number of documents containing term t; Term Frequency-Inverse Document Frequency in Chinese means "term frequency-inverse document frequency", usually abbreviated as TF-IDF, which is a weight calculation method widely used in information retrieval and text mining to evaluate the importance of a term to a document set or one of the documents in a corpus; term frequency (TF) indicates the frequency of a term appearing in a document, while inverse document frequency (IDF) is used to measure the commonness of the term in the entire document collection. A high TF-IDF value of a term in a document means that the term is very common in this document, but not common in the entire document collection, so the term is likely to be a keyword in this document.

[0030] The image features are extracted from the image data using a preset expression, and the preset expression is:

[0031]

[0032] Among them, S(i,j) represents the pixel value of the output feature map of the convolution operation at position (i,j), I(i+m,j+n) is the pixel value of the input image at position (i+m,j+n), K(m,n) is the element of the convolution kernel, and m,n represent the index of the convolution kernel;

[0033] Numerical features are the numerical features of pollutants, site characteristics, environmental factors, etc. obtained through operations such as sensor arrays, spatial interpolation, and on-site measurements.

[0034] The present invention also provides a site pollution control plan generation device, the device comprising:

[0035] A data collection unit, used to collect pollution data of multiple dimensions in the target area, wherein the pollution data at least includes pollutant characteristic data, site characteristic data and environmental factor data;

[0036] A data fusion unit is used to process the pollution data of each dimension respectively, and fuse the pollution data of each dimension after the data processing to obtain fused multimodal data;

[0037] A result generating unit, used for inputting the fused multimodal data into a pre-built solution generating model to obtain a pollution control solution output by the solution generating model;

[0038] The solution generation model is constructed using the historical site pollution data after data processing and the historical governance solutions corresponding to the historical site pollution data.

[0039] The present invention also provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the program.

[0040] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and the computer program implements the steps of the above method when executed by a processor.

[0041] The site pollution control scheme generation method and device provided by the present invention collect pollution data of multiple dimensions of the target area, wherein the pollution data at least includes pollutant characteristic data, site characteristic data and environmental factor data; respectively process the pollution data of each dimension, and fuse the pollution data of each dimension after data processing to obtain fused multimodal data; input the fused multimodal data into a pre-constructed scheme generation model to obtain the pollution control scheme output by the scheme generation model; wherein the scheme generation model is constructed using historical site pollution data after data processing and historical control schemes corresponding to the historical site pollution data.

[0042] In this way, the method and device provided by the present invention utilize massive unstructured data such as typical site pollution control engineering projects, and can effectively convert these unstructured data into structured data through deep learning coding, so that according to retrieval enhancement generation (RAG), prompt engineering and other technologies, the artificial intelligence large model can learn the existing massive engineering cases, repair and control technologies, and multi-source and multi-modal data of pollution on the site, and migrate and apply to form a multi-modal professional intelligent body for site soil and groundwater pollution control, so as to realize intelligent recommendation and iterative optimization of technical solutions, while ensuring the security of sensitive data, data timeliness and accurate decision-making. Thus, through model construction and retrieval, pollution control solutions that can be selected and referenced are generated, solving the technical problem that the generation of site pollution control solutions completely depends on manual experience and on-site investigation. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0044] The structures, proportions, sizes, etc. illustrated in this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with the technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantial technical significance. Any structural modification, change in proportion or adjustment of size shall still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and purposes that can be achieved by the present invention.

[0045] Figure 1 This is one of the flow charts of the method for generating a site pollution control plan provided by the present invention;

[0046] Figure 2 This is the second flow chart of the method for generating a site pollution control plan provided by the present invention;

[0047] Figure 3 This is the third flow chart of the method for generating a site pollution control plan provided by the present invention;

[0048] Figure 4 This is a schematic diagram of the structure of the site pollution control solution generation device provided by the present invention;

[0049] Figure 5 This is a structural block diagram of a computer device provided by the present invention. DETAILED DESCRIPTION

[0050] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] In a specific embodiment, Figure 1 As shown, the method for generating a site pollution control plan provided by the present invention comprises the following steps:

[0052] S110: Collect pollution data of multiple dimensions in the target area, wherein the pollution data at least includes pollutant characteristic data, site characteristic data and environmental factor data; wherein the pollutant characteristic data at least includes the type, concentration and spatial distribution of pollutants; the site characteristic data at least includes soil type, pH value, organic matter content and groundwater distribution; the environmental factor data at least includes temperature, humidity and rainfall data of the polluted site, and these data can be collected by various instruments arranged at the collection nodes.

[0053] S120: Process the pollution data of each dimension respectively, and fuse the pollution data of each dimension after data processing to obtain fused multimodal data; since the collected data involves multiple formats, in order to ensure the uniformity of the data format so as to facilitate input into the subsequent solution generation model, it is necessary to process and fuse the data of each dimension, and the fused data is used as the input data of the subsequent solution generation model.

[0054] S130: Input the fused multimodal data into a pre-built solution generation model to obtain a pollution control solution output by the solution generation model; theoretically, the input pollution control solution can be one or more, displayed in a table, document or other visual form to facilitate reference and screening by staff.

[0055] Among them, the solution generation model is constructed using various types of historical site pollution data after data processing, as well as historical governance solutions corresponding to the historical site pollution data; in the process of model construction, domestic and foreign literature, case studies, project reports, etc. on site pollution control can be collected to form rich unstructured and structured data sets as the external knowledge base of the large model; the collected data is cleaned and labeled, and key features and information are extracted to ensure the quality and consistency of the data; natural language processing (NLP) technology is used, combined with semantic understanding and keyword matching, to construct a large model of site pollution remediation technology based on retrieval enhanced generation, which is the above-mentioned solution generation model.

[0056] During use, the solution generation model can be called up to quickly retrieve relevant technologies and cases in the knowledge base according to user needs. Specifically, according to the needs of site ecological restoration, environmental risk reduction or green and low-carbon technology, the solution generation model uses the coded information obtained after data processing and fusion to cross-modal analysis to identify the complex pollution characteristics in the site, determine the type, source and potential diffusion path of pollutants, and automatically generate the best or better site remediation proposals, intelligently recommend pollution control solutions, such as chemical remediation, physical isolation or biological remediation technology, and adjust the remediation strategy through continuous optimization algorithms to ensure the maximization of site remediation effects. In other words, this method first constructs a multimodal large model for enhanced retrieval (various types of site management materials as a knowledge base, and domestic large models such as Wenxin Yiyan as large models), collects various data on the site that needs to be managed, and then integrates these multimodal data for large model analysis, and finally gives a management plan.

[0057] In some embodiments, Figure 2 As shown in the figure, the process of building a solution generation model specifically includes the following steps:

[0058] S210: Obtain a large amount of historical site pollution data of various types; in specific usage scenarios, when building a solution generation model, first start with extensive and in-depth collection of structured and unstructured data from diversified sources such as typical site pollution control projects, engineering cases, academic literature, professional reports, etc. accumulated worldwide, especially at home and abroad; structured data usually refers to information stored in the form of tables, databases, etc., with clear fields and formats, such as specific parameters of pollution control technology, detailed data of successful cases, specific concentration values ​​of site pollutants, etc.; unstructured data covers a wider range of content, such as project descriptions, case analyses, expert opinions, technical documents and other text information; these data together constitute the basis for analyzing site pollution characteristics and formulating control strategies, and form the structured knowledge base required for large models.

[0059] S220: Clean the historical site pollution data and mark the target information; specifically, data cleaning and preprocessing include deduplication and standardization, text segmentation and stop word removal, etc.; among them, deduplication and standardization can use hash functions or string comparison algorithms (such as Jaccard similarity) to remove duplicate documents; text segmentation can use segmentation tools (such as Jieba segmentation) to decompose text into words or phrases; stop word removal refers to removing common stop words to improve the accuracy of information retrieval. The cleaned data can be expressed as: Cleaned_Data = {xi∈Raw_Data|f(xi)>

[0060] Threshold}, where f(xi) is used to evaluate the relevance of the data.

[0061] S230: Use natural language processing algorithms to extract target features from annotated historical site pollution data, combine semantic understanding and keyword matching to form a site pollution control knowledge base, and build a large model of site pollution remediation technology based on retrieval enhanced generation.

[0062] Specifically, when performing feature extraction and data normalization in step S230, NLP (natural language processing) technology is used, such as word segmentation, part-of-speech tagging, named entity recognition, etc., to accurately extract key features in the text and ensure data quality and consistency; feature extraction is to extract information useful for analysis and prediction from the original data to reduce the dimension and improve model efficiency; data normalization is to convert data of different scales into a unified standard to eliminate model performance problems caused by scale differences; through feature extraction and data normalization, data becomes easier to process, which improves the accuracy and convergence speed of the model. Keyword matching quickly locates key information by directly looking up predefined terms or phrases, and is suitable for processing clear and structured content: Applicable scenarios: technology names (such as "chemical immobilization"), pollutant names (such as "lead", "benzene"). Implementation method: based on regular expression or string matching technology. Semantic understanding uses a large model to deeply analyze unstructured text and extract implicit information. It is suitable for processing vague and indirect content: Applicable scenarios: descriptive information about contaminated sites (such as "high concentrations of heavy metals in groundwater"), technical applicability conditions (such as "suitable for high pH soil"). Implementation method: Combine named entity recognition (NER), dependency syntax analysis and other technologies to obtain contextual semantics through a large model.

[0063] In specific usage scenarios, feature extraction can use TF-IDF (Term Frequency-Inverse Document Frequency):

[0064]

[0065] Among them, TF(t,d) is the frequency of term t in document d, N is the total number of documents, and DF(t) is the number of documents containing term t. Term Frequency-Inverse Document Frequency in Chinese means "term frequency-inverse document frequency", usually abbreviated as TF-IDF. It is a weight calculation method widely used in information retrieval and text mining to evaluate the importance of a word to a document set or one of the documents in a corpus; term frequency (TF) indicates the frequency of a word appearing in a document, while inverse document frequency (IDF) is used to measure the commonness of the word in the entire document collection. A high TF-IDF value of a word in a document means that the word is very common in this document, but not common in the entire document collection, so the word is likely to be a keyword in this document.

[0066] Word embedding can use Word2Vec, which maps each word into a vector of fixed dimension. The formula is as follows:

[0067]

[0068] Among them, w t is the central word, w i Context words, n is the size of the context window, and is used to predict context words (context words) based on the current word (center word). Given a sentence, the model hopes to maximize the probability of predicting context words given a center word. Word2Vec is a computational model used to generate word vectors, that is, the representation of words in a continuous vector space. This computational model was launched by Google in 2013 and is an efficient distributed word representation method that can capture the semantic relationship between words. Word2Vec has two architectures: continuous bag of words (CBOW) and Skip-Gram. For details, please refer to the existing technology and will not be described in detail.

[0069] Specifically, data normalization can be achieved using the following formula:

[0070]

[0071] x ′ represents the data after normalization, and x represents the data before normalization. This formula is used to normalize the data x to the range of [0,1], where min(X) and max(X) are the minimum and maximum values ​​of the data set x, respectively. This process ensures that data from different sources and dimensions can be compared and analyzed on a unified scale.

[0072] The solution generation model can be understood as a large model based on the retrieval model. Considering the security of the large model, a domestic large model is used. The enhanced retrieval model can be implemented using BM25 (Best Matching 25), and the specific expression is:

[0073]

[0074] Among them, BM25(q,d) represents the relevance score between document d and query q, q represents query (consisting of one or more query terms), d represents document (candidate document in the retrieval system), t i represents the i-th term in the query, f(t i ,d) is the word t i frequency in document d, where |d| is the length of the document, is the average length of the document, IDF(t i ) represents the term t i The inverse document frequency (measure of the importance of terms), k and b are adjustment parameters, and their value ranges are [1.2, 2.0] and [0, 1] respectively.

[0075] The retrieval module uses similarity calculation (such as cosine similarity) to retrieve relevant documents from the knowledge base. The expression of phase velocity is:

[0076]

[0077] Among them, CosineSimilarity(A,B) represents AB cosine similarity, A represents the content of the knowledge base, B represents the content of the query, that is, the venue data, ||A|| and ||B|| represent the modulus length of the vector.

[0078] The generation module is used to generate responses based on the retrieval results, using beam search (heuristic search algorithm) or top-k sampling to optimize the diversity and accuracy of generation.

[0079] Based on the above modular retrieval process, the prompt words required by the large model are automatically generated, so as to output the remediation technology for specific contaminated sites. The expression of the prompt word project is:

[0080] Prompt=Template(Variables)

[0081] Among them, Prompt refers to the instructions or questions that the generative model needs to input to guide the model output, Template (Variables) represents the format or template that defines the prompt words (for example, "Provide remediation suggestions for the following sites: {site conditions}"); Variables represents dynamic variables, which represent the input data provided by the user (such as site conditions, pollutant characteristics).

[0082] Furthermore, the characteristic data of the pollutants in the site to be treated are processed, and then the following steps are also included:

[0083] Determine the concentration levels of target pollutants in soil and water samples based on extensive collection of soil and water samples in the target area;

[0084] The concentration distribution map of the target pollutant is drawn by using the sampling data and the spatial interpolation method, and the spatial distribution characteristics and migration trends of the target pollutant are determined based on the concentration distribution map.

[0085] In a specific use scenario, in the process of collecting and processing multimodal pollutant sample data of a contaminated site, multidimensional pollution data of the site, such as pollutant characteristics, site characteristics and environmental factors, are collected through IoT devices. The pollutant characteristics include the type, concentration and spatial distribution of pollutants, the site characteristics include the type, pH value, organic matter content and groundwater distribution of soil, and the environmental factors include the temperature, humidity and rainfall data of the contaminated site. In the process of collecting and transmitting multimodal data of the site, multidimensional pollution data of the site, including pollutant characteristics, site characteristics and environmental factors, are collected and transmitted in real time through on-site collection, sensor arrays and IoT devices. The data types of the pollution data include at least text data, image data and sensor reading data.

[0086] Specifically, there are many types of pollutants, including heavy metals, organic pollutants, and petroleum pollutants, and their concentrations and spatial distributions vary. In order to accurately grasp this information, in this embodiment, the sampling methods used include grid point distribution method and random point distribution method, etc., and soil and water samples are widely collected in the contaminated site; these samples are then sent to the laboratory for in-depth analysis to determine the specific concentration level of the pollutants; combined with the geographic information system (GIS), using sampling data and spatial interpolation methods, a pollutant concentration distribution map can be drawn, thereby clearly showing the spatial distribution characteristics and potential migration trends of pollutants.

[0087] In terms of site characteristics, this example focuses on key factors such as soil type, pH value, organic matter content, and groundwater distribution; through careful on-site observation and professional laboratory analysis, the soil type and texture of the contaminated site are accurately determined; at the same time, representative soil samples are collected and their pH value and organic matter content are accurately measured to fully understand the basic chemical properties of the soil. In addition, this example makes full use of geological exploration data and hydrogeological survey data to deeply analyze the distribution range, flow direction, and flow rate of groundwater, and then scientifically evaluate the migration risk of pollutants in groundwater.

[0088] Environmental factors also have an important impact on the diffusion and treatment effect of pollutants. In this embodiment, advanced meteorological monitoring equipment is equipped to regularly record the temperature and humidity data of the contaminated site, and deeply analyze the potential impact of these data on the migration and transformation of pollutants. At the same time, historical rainfall data is widely collected, and the specific effects of rainfall on the scouring, migration and diffusion of pollutants are deeply analyzed to predict the changing trend of pollutants under future rainfall conditions. In addition, according to the specific circumstances, this embodiment also fully considers other environmental factors that may affect the remediation effect, such as wind direction, wind speed and solar radiation.

[0089] In this embodiment, historical remediation records generally include remediation plans and technologies, pollutant characteristics, remediation effect evaluation and monitoring data; among them, remediation plans and technologies specifically refer to past remediation methods (such as biological remediation, chemical remediation, physical isolation, etc.) and the technical equipment and materials used; pollutant characteristics include data such as pollutant types, concentrations, diffusion paths, etc. at historical remediation sites; remediation effect evaluation includes changes in pollutant concentrations after remediation, remediation time, cost, environmental impact, etc.; monitoring data refers to monitoring data during and after the remediation process, including pollutant concentrations, environmental conditions, changes in site characteristics, etc.; data is converted into a usable data set through data cleaning, labeling and structured processing.

[0090] In order to ensure the real-time and integrity of the data, this embodiment uses a cloud platform to efficiently transmit and store multimodal data; the collected data covers multiple aspects such as soil, groundwater, and air quality, as well as sensor electrical signal data, image monitoring data, and historical governance records. Through the powerful functions of the cloud platform, we can achieve centralized management and rapid access to these data, providing strong support for subsequent pollution analysis and governance work.

[0091] In some embodiments, the pollution data of each dimension after data processing is fused, specifically including:

[0092] The text data is encoded using a word embedding algorithm to obtain an embedding vector corresponding to the historical governance record text data of the site to be governed;

[0093] extracting image features from the image data;

[0094] Extract pollutant and environmental numerical characteristics;

[0095] The embedding vector and the image feature are passed through a data fusion algorithm to obtain a fusion feature output by the deep learning model; specifically, through deep learning coding (word coding, image feature extraction, numerical normalization, etc.) technology and a data fusion algorithm, multimodal data, such as electrical signal data of sensors such as soil, groundwater, and air quality, image monitoring, and historical pollution control record text data of the required control site are fused, and the multimodal data is transmitted and processed through a cloud platform to form a unified data model to provide support for subsequent pollution analysis and decision-making.

[0096] Based on the pre-processed multimodal data, this embodiment constructs a cross-modal contaminated site data fusion model to fully utilize the advantages of deep learning technology to accurately identify the types, sources and diffusion paths of pollutants in complex contaminated sites. Through cross-modal analysis, the model can comprehensively consider multiple factors, such as soil properties (including soil type, pH value, organic matter content, etc.), environmental conditions (such as temperature, humidity, rainfall, etc.) and treatment technology characteristics, and conduct a comprehensive and in-depth analysis.

[0097] Text data encoding can use Word2Vec, which maps each word into a vector of fixed dimension. The formula is as follows:

[0098]

[0099] Among them, w t is the central word, w i Context words, n is the size of the context window, and is used to predict context words (context words) based on the current word (center word). Given a sentence, the model hopes to maximize the probability of predicting context words given a center word. Word2Vec is a computational model used to generate word vectors, that is, the representation of words in a continuous vector space. This computational model is an efficient distributed word representation method that can capture the semantic relationship between words. Word2Vec has two architectures: continuous bag of words (CBOW) and Skip-Gram. For details, please refer to the existing technology and will not be described in detail.

[0100] The image features are extracted using a convolutional neural network, and the specific expression is:

[0101]

[0102] Among them, S(i,j) represents the pixel value of the output feature map of the convolution operation at position (i,j), I(i+m,j+n) is the pixel value of the input image at position i+m,j+n, K(m,n) is the element of the convolution kernel, and m, n represent the index of the convolution kernel.

[0103] The expression of cross-modal data fusion is:

[0104] Y=g(X1,X2,…,X n ;θ)

[0105] Among them, Y is the model output (multimodal data fusion vector), X i is the data of the i-th modality (such as on-site images, historical texts, etc.), the value range of i is 1-n, g is the deep learning model, θ is the model parameter, and the formula represents how to integrate the features of multiple modal data.

[0106] Furthermore, in the process of making intelligent decisions and recommending pollution control solutions, based on the needs, such as site ecological restoration, environmental risk reduction, green and low-carbon technology, etc., relying on the multimodal large model of site remediation and control technology, our system has demonstrated outstanding intelligence in the field of site pollution remediation. The workflow of the large model begins with the accurate grasp of user needs and the in-depth analysis of pollutant data collected on site. In the solution generation model based on the construction, according to the actual multimodal data of the site, through the prompt word engineering, the optimal control solution is intelligently selected from the massive control technology library, so as to output the optional control solution to assist in the generation of solutions. The generated solutions cover a variety of technical means such as physical remediation, chemical remediation, and biological remediation, as well as a comprehensive remediation system integrating multiple technologies. Physical remediation may include soil excavation, leaching, solidification / stabilization and other methods; chemical remediation may involve chemical reactions such as oxidation, reduction, and extraction; biological remediation uses the metabolic activities of microorganisms or plants to degrade pollutants, as well as an automatically derived integrated solution system. The control solutions recommended by the model are not only based on the current pollution situation, but also fully consider the future trend of change. Through continuous optimization of algorithms, the remediation strategy can be continuously iterated to adapt to changes in pollution conditions. This dynamic adjustment capability ensures the maximization of the remediation effect and avoids poor governance results caused by rigid solutions.

[0107] In actual operation, the governance plan generation method can be implemented with the system as the carrier, and the system can provide users with a detailed governance plan report, including the specific steps of the recommended technology, the required materials, the expected effects, and the possible risks and countermeasures. At the same time, the system also supports users to adjust and optimize the plan according to the actual situation to ensure that the governance plan finally implemented is both scientific and feasible.

[0108] Further, in order to achieve continuous monitoring and adaptive optimization, in the implementation stage of site pollution remediation, the system on which this embodiment relies demonstrates a powerful real-time monitoring and optimization capability. Through Internet of Things devices, such as sensor networks, drone inspections, remote monitoring cameras, etc., the system can continuously collect the latest data of the site, including key indicators such as soil pollution concentration, groundwater quality, air quality, temperature, and humidity. These data are transmitted to the system background in real time, providing strong support for the monitoring of the remediation effect. The system uses advanced algorithms and models to conduct in-depth analysis of real-time monitoring data to evaluate the effectiveness of remediation measures. By comparing the data changes before and after the remediation, the system can intuitively display the remediation effect, such as the degree of reduction in pollutant concentration, the improvement of water quality and air quality, etc. This real-time feedback mechanism ensures that the remediation process is always in a controllable state, which is convenient for timely discovery and resolution of problems. More importantly, the system does not just stay at the monitoring level, but takes new monitoring data as input to continuously iterate and optimize the remediation plan. Through data-driven decision support, the system can automatically adjust the remediation strategy, such as changing the parameters of physical remediation, adjusting the dosage of chemical reagents, optimizing the conditions of biological remediation, etc., to adapt to changes in pollution conditions and the needs of remediation effects. This dynamic optimization process ensures that the restoration plan always remains in the best state, which not only improves the restoration efficiency but also reduces the secondary impact on the environment.

[0109] During the implementation of the remediation, the system will also take into account a variety of factors, such as weather conditions, soil type, groundwater flow direction, etc., to develop a more accurate remediation strategy. By comprehensively considering these factors, the system can predict and respond to potential challenges and risks to ensure the safety and reliability of the remediation process. Ultimately, through continuous monitoring, real-time feedback and data iteration, the system can ensure that the efficiency and effectiveness of site remediation reach the optimal level; at the same time, by dynamically optimizing the pollution control process, we minimize the secondary impact on the environment and achieve green and sustainable remediation goals. In order to ensure data security and privacy, the system uses strict data encryption technologies, such as data desensitization and permission control mechanisms, especially for the processing of sensitive data, to ensure the timeliness and confidentiality of the data.

[0110] For ease of understanding, the implementation process of the method provided by the present invention is briefly described below using a specific usage scenario as an example.

[0111] like Figure 3As shown in the figure, a chemical park has serious soil and groundwater pollution due to long-term industrial production activities. The main pollutants include heavy metals (such as chromium and copper) and organic chemicals (such as benzene and ketone solvents). The presence of these pollutants not only affects the soil quality, but also seeps into the groundwater, posing a potential threat to the surrounding ecological environment and the health of residents. To this end, it is necessary to implement contaminated site remediation and adopt an intelligent decision-making method based on a multimodal large model to guide the formulation of remediation plans.

[0112] Pollution monitoring and data collection: Through a combination of drone remote sensing, ground sampling and sensor arrays, comprehensive pollutant sample data was obtained for polluted sites within the park; the data included environmental factors such as heavy metal content in soil and groundwater, chemical concentration, soil pH, temperature and humidity, rainfall, and the first phase of governance records, to fully understand the types, concentrations and spatial distribution of pollutants.

[0113] Analysis of pollutant characteristics and remediation targets: The analysis results show that the concentrations of heavy metals such as chromium and copper in the park are seriously exceeded, and there is large-scale pollution by benzene and ketone chemical solvents; through the integration of data such as soil type and groundwater flow conditions, the model determines the mobility and impact range of pollutants, and further defines the area to be remediated.

[0114] Intelligent decision-making and formulation of remediation plans: After inputting pollutant characteristics, site characteristics and environmental data into the multimodal big model with enhanced retrieval in the knowledge base (this embodiment takes the domestic reasoning big model Tiangong 4.0 (o1) as an example), the big model quickly analyzes and proposes remediation goals and methods. Based on the recommended plan generated by the model, the remediation team formulated the following specific remediation plan and gave relevant parameters, such as pollutant concentration thresholds, analysis of possible pollution diffusion paths, parameters for the time and space addition of agents, treatment cycles, etc.

[0115] Restoration of heavy metal polluted areas;

[0116] Soil leaching method: using high-efficiency soil leaching technology to remove heavy metal ions such as chromium and copper from the soil through water solvent;

[0117] Solidification and stabilization technology: Use a solidifying agent to fix the remaining heavy metals in the soil to prevent them from migrating and spreading again;

[0118] Regional remediation of organic pollutants;

[0119] Biodegradation technology: by introducing specific microorganisms to promote the degradation of benzene and ketone solvents, reducing their harm to the environment and human body;

[0120] Chemical oxidation technology: Add oxidants to promote the transformation of organic pollutants and enhance the degradation effect. Remediation of complex polluted areas: The remediation plan for complex polluted areas adopts a combination of various technical means according to the characteristics of different pollutants, aiming to minimize the impact of pollutants on soil and groundwater, achieve effective removal of pollutants and ecological restoration of the site.

[0121] Remediation of areas where heavy metals and organic pollutants coexist, physical and chemical remediation methods: soil leaching method combined with chemical oxidation method: while dealing with heavy metal pollution, soil leaching method is used to remove soluble heavy metals, and chemical oxidation method can be used to degrade accompanying organic pollutants. The combination of these two can effectively reduce the concentration of pollutants in a short period of time and enhance the extensiveness of remediation; modified soil solidification method: for heavy metal residues, solidification stabilization technology can be used. The modified soil solidification method fixes harmful substances in the soil by adding specific chemicals to prevent the migration of pollutants. The organic matter that may be released during the solidification process is further degraded by chemical oxidation technology.

[0122] Restoration of areas where organic pollutants and nutrients coexist, bioremediation technology and nutritional regulation: Biodegradation and nutritional supplementation method: For the degradation of organic pollutants, specific microbial groups are used for bioremediation, and the growth and degradation efficiency of microorganisms are promoted by adjusting the nutrients in the soil (such as nitrogen, phosphorus, etc.). The restoration effect can be further improved by optimizing environmental conditions. Microbial induced sedimentation method: Under certain circumstances, the degradation of organic pollutants in the soil can be promoted through the action of microorganisms. At the same time, the byproducts produced by microorganisms during the degradation process can help solidify heavy metals and form a more stable complex pollution environment.

[0123] Multi-pollutant alternating remediation technology, phased remediation method:

[0124] Phase 1: Pollutant pretreatment: For complex pollution, physical and chemical methods are first used to remove pollutants. For example, soil leaching is used to reduce heavy metal concentrations, or surfactant extraction is used to remove organic pollutants. The purpose of this phase is to reduce pollutant concentrations through simple and efficient methods to create conditions for subsequent bioremediation.

[0125] Phase 2: Combination of bioremediation and stabilization: After pretreatment, bioremediation technology is used to treat the microbial degradation of organic pollutants, while stabilization technology is used to fix heavy metals in the soil. The focus of this phase is to improve the durability of the remediation and reduce the possibility of pollutant migration.

[0126] Phase 3: Synergy of ecological restoration and composite restoration system: Restore the ecological function of the site through measures such as phytoremediation and soil improvement, and ensure long-term restoration through intelligent monitoring and adjustment.

[0127] Remediation process and monitoring: During the remediation process, monitoring technology is used to track the remediation effect in real time, including key parameters such as pollutant concentration, soil and groundwater quality; based on the monitoring results, the remediation team can dynamically adjust the remediation plan to ensure that the remediation effect achieves the expected goal.

[0128] Restoration results: By adopting a multimodal intelligent decision-making method, the remediation of contaminated sites in the chemical park has achieved remarkable results; the remediation team can quickly and accurately determine the remediation goals and plans, avoiding the inefficiency and blindness that may occur in traditional remediation methods; due to the full integration of multi-source data of environmental factors from remote sensing, ground sampling, and sensor arrays, and based on the retrieval and reasoning of large models in existing technology libraries, the decision results are more accurate, effectively improving the remediation effect and reducing the remediation cost.

[0129] The successful application of this project verified the effectiveness of the intelligent decision-making method for site pollution remediation based on a multimodal large model. Through comprehensive data analysis and model prediction, the remediation work became more scientific and efficient, providing experience and methods that can be used as reference for future remediation of contaminated sites.

[0130] In the above specific implementation, the method provided by the present invention utilizes massive unstructured data such as typical site pollution control engineering projects, and can effectively convert these unstructured data into structured data through deep learning coding, so that according to retrieval enhancement generation (RAG), prompt engineering and other technologies, the artificial intelligence large model learns the existing massive engineering cases, repair and control technologies, and multi-source and multi-modal data of pollution on the site, and migrates and applies them to form a multi-modal professional intelligent body for site soil and groundwater pollution control, so as to realize intelligent recommendation and iterative optimization of technical solutions, while ensuring the security of sensitive data, data timeliness and accurate decision-making. Thus, through model construction and retrieval, pollution control solutions that can be selected and referenced are generated, which solves the technical problem that the generation of site pollution control solutions completely depends on manual experience and on-site investigation.

[0131] In addition to the above method, the present invention also provides a site pollution control plan generation device, such as Figure 4 As shown, the device comprises:

[0132] A data collection unit 410 is used to collect pollution data of multiple dimensions in a target area, wherein the pollution data at least includes pollutant characteristic data, site characteristic data and environmental factor data;

[0133] The data fusion unit 420 is used to process the pollution data of each dimension respectively, and fuse the pollution data of each dimension after the data processing to obtain fused multimodal data;

[0134] A result generating unit 430 is used to input the fused multimodal data into a pre-built solution generating model to obtain a pollution control solution output by the solution generating model;

[0135] The solution generation model is constructed using the historical site pollution data after data processing and the historical governance solutions corresponding to the historical site pollution data.

[0136] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 5 As shown. The computer device includes a processor, a memory and a network interface connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a model prediction. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The model prediction of the computer device is used to store static information and dynamic information data. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the steps in the above method embodiment are implemented.

[0137] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0138] Corresponding to the above embodiment, the embodiment of the present invention further provides a computer storage medium, which contains one or more program instructions, wherein the one or more program instructions are used to execute the above method.

[0139] The present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the above method.

[0140] In the embodiment of the present invention, the processor may be an integrated circuit chip having the ability to process signals. The processor may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0141] The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present invention can be directly embodied as a hardware decoding processor for execution, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The processor reads the information in the storage medium and completes the steps of the above method in combination with its hardware.

[0142] The storage medium may be a memory, which may be, for example, a volatile memory or a nonvolatile memory, or may include both volatile and nonvolatile memory.

[0143] Among them, the non-volatile memory can be a read-only memory (ROM), a programmable ROM (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory.

[0144] The volatile memory may be a random access memory (RAM) which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DRRAM).

[0145] The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0146] Those skilled in the art will appreciate that in one or more of the above examples, the functions described in the present invention can be implemented using a combination of hardware and software. When software is used, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any media that facilitates the transmission of computer programs from one place to another. Storage media can be any available media that can be accessed by general or special-purpose computers.

[0147] The above specific implementation methods further illustrate the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific implementation methods of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solutions of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for generating a site pollution control plan, characterized in that: The method comprises: Collecting pollution data of multiple dimensions in the target area, wherein the pollution data at least includes pollutant characteristic data, site characteristic data and environmental factor data; The pollution data of each dimension are processed respectively, and the pollution data of each dimension after the data processing are fused to obtain fused multimodal data; Inputting the fused multimodal data into a pre-built solution generation model to obtain a pollution control solution output by the solution generation model; The solution generation model is constructed using the historical site pollution data after data processing and the historical governance solutions corresponding to the historical site pollution data.

2. The method for generating a site pollution control plan according to claim 1, characterized in that: Building a solution generation model specifically includes: Obtain a large amount of historical site contamination data of various types; Clean the historical site pollution data and mark the target information; By using natural language processing algorithms, target features are extracted from the annotated historical site pollution data. Combined with semantic understanding and keyword matching, a large model of site pollution remediation technology based on retrieval enhanced generation is constructed.

3. The method for generating a site pollution control plan according to claim 1, characterized in that: The pollutant characteristic data at least includes the type, concentration and spatial distribution of the pollutant; and / or, The site characteristic data includes at least soil type, pH value, organic matter content and groundwater distribution; and / or, The environmental factor data at least include temperature, humidity and rainfall data of the contaminated site.

4. The method for generating a site pollution control plan according to claim 1, characterized in that: Data processing of pollutant characteristic data, followed by: Determine the concentration levels of target pollutants in soil and water samples based on extensive collection of soil and water samples in the target area; The concentration distribution map of the target pollutant is drawn by using the sampling data and the spatial interpolation method, and the spatial distribution characteristics and migration trends of the target pollutant are determined based on the concentration distribution map.

5. The method for generating a site pollution control plan according to claim 1, characterized in that: The data types of the pollution data include at least text data, image data and sensor reading data.

6. The method for generating a site pollution control plan according to claim 5, characterized in that: The pollution data of each dimension after data processing is integrated, including: The text data is encoded using a word embedding algorithm to obtain an embedding vector corresponding to the historical governance record text data of the site to be governed; extracting image features from the image data; Extracting pollutant and environmental numerical features; applying a data fusion algorithm to the embedded vector and the image features to obtain fusion features output by the deep learning model.

7. The method for generating a site pollution control plan according to claim 6, characterized in that: Document feature extraction uses TF-IDF: Where TF(t,d) is the frequency of word t in document d, N is the total number of documents, and DF(t) is the number of documents containing word t; The image features are extracted from the image data using a preset expression, and the preset expression is: Among them, S(i,j) represents the pixel value of the output feature map of the convolution operation at position (i,j), I(i+m,j+n) is the pixel value of the input image at position i+m,j+n, K(m,n) is the element of the convolution kernel, and m,n represent the index of the convolution kernel; Numerical features are the numerical features of pollutants, site characteristics, environmental factors, etc. obtained through operations such as sensor arrays, spatial interpolation, and on-site measurements.

8. A site pollution control plan generation device, characterized in that: The device comprises: A data collection unit, used to collect pollution data of multiple dimensions in the target area, wherein the pollution data at least includes pollutant characteristic data, site characteristic data and environmental factor data; A data fusion unit is used to process the pollution data of each dimension respectively, and fuse the pollution data of each dimension after the data processing to obtain fused multimodal data; A result generating unit, used for inputting the fused multimodal data into a pre-built solution generating model to obtain a pollution control solution output by the solution generating model; The solution generation model is constructed using the historical site pollution data after data processing and the historical governance solutions corresponding to the historical site pollution data.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images